Dear SPM experts,
I want to perform a very simple connectivity analysis in SPM8. Basically, I just want to know which brain regions correlate with a seed region over the whole experiment. So what I do is the following:
1) for each participant, I extract the eigenvariate of the seed region (a 6mm sphere in the IFC) using the VOI time series extraction method provided by SPM
2) I then use the extracted time course of activation as a regressor in the GLM (in the batch I enter the timecourse in the field "Multiple Regressors")
3) to see how the seed region correlates with other regions of the brain, I then calculate contrast images for each participant (basically I just have [1] in the contrast vector because in the most simple case the seed region activation is my only regressor, although I also tried other models for which I entered experimental conditions and motion parameters as additional regressors)
4) on the obtained contrast images I then perform a second-level analysis
Does this sound like a reasonable procedure?
The problem is that more or less the whole brain appears to correlate with my seed region. Only if I set the significance threshold extremely low (0.0000001, FWE-corrected) do I get reasonable activations that don't include the whole brain. So what's wrong with the analysis? Do I maybe have to do some special kind of filtering (I use a standard 1/128 s highpass filter in the GLM)? How about global normalisation?
I really appreciate your input!
Thanks a lot,
Moritz
|